Security Processing of Street View Image Information

In the era of cyber-physical-social service convergence, the applications of image scenes have become more and more complex, leading to serious image security problems. In order to realize the security processing of street scene image information, this paper designs a single-level face and license plate detection model based on YOLO v5x to realize the automatic recognition and detection of faces and license plates in street scenes, and then erases the recognition region to realize the desensitization of street scene image information at the pixel level. On this basis, an image restoration model based on atrous deformable convolution is proposed to accomplish the restoration of street scene images. In this paper, we use the self-made Hong Kong street view dataset for model training. Compared with the YOLO series of target detection models and the classical image complementation model, better performance is achieved, and the secure processing of the information in the actual scene map of Hainan is also realized.